Bidirectional Associations Between Circulating Polyunsaturated Fatty Acids and Female Reproductive Endocrine-Related Diseases: A Mendelian Randomization Study

Female reproductive endocrine disorders, including endometriosis, infertility, polycystic ovary syndrome (PCOS), premenstrual syndrome (PMS), and premature ovarian insufficiency (POI), are among the most prevalent and debilitating conditions affecting women of reproductive age worldwide. 1 Collectively, these disorders impact an estimated 10–20% of women globally, posing significant challenges to fertility, mental health, and overall quality of life. 2 Endometriosis alone is associated with an annual economic burden exceeding $12,000 per patient in direct healthcare costs. 3 Similarly, PCOS is associated with an increased risk of metabolic syndrome, type 2 diabetes, and cardiovascular disease, reflecting the broader systemic implications of these disorders. 4 Despite advances in diagnostic and symptomatic management strategies, the underlying etiologies of these conditions remain incompletely understood, involving complex interactions between genetic predisposition, endocrine dysregulation, immune dysfunction, and environmental factors. 5–7 There is an urgent need to identify modifiable biomarkers that could pave the way for novel prevention and therapeutic strategies. Polyunsaturated fatty acids (PUFAs), comprising omega-3 (eg., docosahexaenoic acid [DHA]) and omega-6 (eg., linoleic acid) subtypes, are emerging as key modulators in reproductive physiology and pathophysiology. 8 Omega-3 PUFAs are known to exert anti-inflammatory effects by inhibiting cyclooxygenase-2 activity and reducing prostaglandin E2 synthesis, mechanisms that may limit the proliferation of ectopic endometrial tissue in endometriosis. 9 , 10 Conversely, omega-6 PUFAs serve as precursors for pro-inflammatory eicosanoids such as leukotrienes and thromboxanes, which may exacerbate ovarian hyperandrogenism and insulin resistance in PCOS. 11 Observational studies have highlighted inverse associations between dietary omega-3 intake and endometriosis risk, with a pooled odds ratio (OR) of 0.78 (95% confidence interval [CI]: 0.65–0.94) in meta-analyses. 12 However, the evidence for omega-6 PUFAs remains inconclusive, with some studies suggesting null or even protective effects. These discrepancies underscore the need for robust methodologies that can disentangle causality from confounding. Although several clinical trials have demonstrated that supplementation of PUFAs (such as omega-3s) can alleviate PMS, the dose and durations of supplementation varied. 13 , 14 In contrast, another study from Houghton et al suggested that individual PUFAs were not associated with risk of PMS. 15 And to best of our knowledge, there was no study have investigated the relationship between PUFAs and POI. Importantly, circulating PUFA levels should not be interpreted simply as direct surrogates of dietary intake. Although plasma and erythrocyte PUFA concentrations provide objective biomarkers of PUFA status and partially reflect habitual dietary intake, they are also influenced by endogenous metabolic processes, including intestinal absorption, hepatic lipid metabolism, fatty acid desaturation and elongation, β-oxidation, adipose tissue storage and mobilization. 16–20 Meanwhile, body mass index (BMI) has been reported to determine the plasma long-chain PUFA response to dietary fat manipulation. 21 Since the widely acknowledged association between BMI and metabolic dysfunction with female reproductive endocrine disorders, 22 , 23 conventional observational studies are vulnerable to confounding by metabolic factors. In addition, reliance on self-reported dietary intake introduces recall bias and measurement error, while correlations between self-reported PUFA intake and circulating PUFA biomarkers are often only moderate, indicating substantial uncertainty in traditional nutritional epidemiology. 24 , 25 In addition, candidate-gene and polymorphism association studies, such as recent evidence on ADIPOQ variants in PCOS, suggest that sequence variations may contribute to PCOS-related phenotypes, although reported SNP–disease associations remain heterogeneous across populations and genetic loci. 26 To address these limitations, Mendelian randomization (MR) represents a robust epidemiological approach that employs genetic variants as instrumental variables (IVs) to proxy lifelong exposure levels. 27 By mimicking the principles of randomized controlled trials, MR reduces the influence of confounding and reverse causation, enabling robust causal inference. 28 The validity of MR findings hinges on three core assumptions: (1) the IVs are robustly associated with the exposure (F-statistic > 10); (2) the IVs are independent of confounders of the exposure-outcome relationship; and (3) the IVs influence the outcome exclusively through the exposure, with no horizontal pleiotropy. 29 , 30 MR has been instrumental in elucidating causal links between lipid traits and cardiometabolic outcomes, 31 yet its application to reproductive endocrine disorders is still in its infancy. To date, no MR study has comprehensively evaluated the bidirectional causal relationships between circulating PUFA subtypes and female reproductive disorders, leaving critical mechanistic gaps unaddressed. This study seeks to fill these gaps through a two-sample MR design, integrating data from large-scale genome-wide association studies (GWAS) on circulating PUFAs and female reproductive disorders. Specifically, we aim to provide novel insights into the role of PUFAs as mediators of reproductive pathophysiology, advancing our understanding of their potential as modifiable targets for intervention. By addressing key limitations in prior observational research and leveraging robust statistical frameworks, this work has the potential to inform precision nutrition and lifestyle interventions tailored to high-risk populations.

An additional file summarizes the GWAS data sources for the PUFA-related traits and the five female reproductive endocrine disorders investigated ( Table S1 ). The exposures, including circulating levels of DHA, linoleic acid, total omega-3, total omega-6, the omega-6/omega-3 ratio, PUFA, and the PUFA to total fatty acids ratio, originate from a European population GWAS (N=115,006).The outcomes of interest were endometriosis, female infertility, PCOS, premenstrual syndrome, and POI, each obtained from distinct large-scale GWAS consortia or FinnGen releases. This dataset provides a comprehensive foundation for forward Mendelian randomization (MR) analyses to test the causal roles of circulating PUFA phenotypes, as well as for reverse MR to explore the potential impact of these disorders on PUFA levels. In the forward MR, each exposure was instrumented by SNPs that reached genome-wide significance ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${\mathrm{P}} < 5 \times {10^{ - 8}}$\end{document} ). An additional file details the number of original, LD-pruned, and palindromic-removed SNPs, along with the resultant F-statistics and proportion of variance explained ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${R^2}$\end{document} ) ( Table S2 ). For instance, 34–43 SNPs were retained for DHA, each set explaining roughly 4.09% to 4.14% of the phenotypic variance, with F-values above 100 in most cases, suggesting robust instruments. Similarly, 41–46 instruments were identified for linoleic acid ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${R^2},{\ }5.12{\mathrm{\% - }}5.65{\mathrm{\% }}$\end{document} ) and 49–57 instruments for total omega-6 ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${R^2},{\ }5.64{\mathrm{\% - }}6.20{\mathrm{\% }}$\end{document} ). This table also indicates if any outlier SNPs were detected by MR-PRESSO or Radial MR, such as rs115478735 for linoleic acid, total omega-6, and PUFA in the endometriosis analysis. No substantial issues with weak instrument bias were observed, as all F-statistics exceeded the conventional threshold of 10. Before estimating causal effects, we checked for potential pleiotropy and heterogeneity among the genetic instruments. Table 1 reports the MR-Egger intercept for each exposure–outcome pair (to detect horizontal pleiotropy), the MR-Steiger direction test, and Cochran’s Q statistics for both MR-Egger and IVW models (to evaluate between-SNP heterogeneity). In nearly all comparisons, the MR-Egger intercept did not reach statistical significance, implying that unbalanced horizontal pleiotropy was not a major issue. The MR-Steiger tests consistently confirmed the hypothesized direction of causality (exposure → outcome). However, Cochran’s Q test revealed notable heterogeneity among the instrumental variables (IVs) for DHA, Total omega-3, Total omega-6, the ratio of omega-6 to omega-3, and PUFA with respect to Polycystic Ovarian Syndrome. This finding prompted the use of a multiplicative random-effects IVW model rather than a fixed-effects model. Table 1 The Pleiotropic and Heterogeneity Test Outcome and Exposure Horizontal Pleiotropic Test Casual Direction Test Heterogeneity Test Egger Intercept P MR Steiger P MR Egger Q P IVW Q P Endometriosis  DHA −0.008 0.109 TRUE <0.001 42.725 0.397 45.516 0.328  Linoleic acid 0.009 0.242 TRUE <0.001 48.659 0.291 50.215 0.274  Total omega-3 0.002 0.693 TRUE <0.001 53.551 0.237 53.730 0.264  Total omega-6 0.002 0.701 TRUE <0.001 61.800 0.246 61.967 0.272  Ratio of omega-6 to omega-3 0.002 0.818 TRUE <0.001 38.085 0.097 38.159 0.119  PUFA −0.001 0.894 TRUE <0.001 61.555 0.197 61.576 0.223  Ratio of PUFA to total fatty acids −0.013 0.076 TRUE <0.001 54.238 0.117 58.398 0.072 Female infertility  DHA 0.005 0.444 TRUE <0.001 50.918 0.138 51.661 0.146  Linoleic acid 0.015 0.055 TRUE <0.001 28.611 0.965 32.500 0.918  Total omega-3 0.007 0.262 TRUE <0.001 63.047 0.059 64.775 0.054  Total omega-6 0.004 0.522 TRUE <0.001 50.494 0.647 50.909 0.667  Ratio of omega-6 to omega-3 0.000 0.956 TRUE <0.001 20.312 0.853 20.315 0.883  PUFA 0.002 0.773 TRUE <0.001 53.845 0.442 53.931 0.477  Ratio of PUFA to total fatty acids −0.011 0.149 TRUE <0.001 51.921 0.165 54.522 0.133 Polycystic ovarian syndrome  DHA 0.007 0.615 TRUE <0.001 48.593 0.030 48.985 0.036  Linoleic acid 0.016 0.319 TRUE <0.001 52.688 0.071 54.067 0.068  Total omega-3 0.010 0.399 TRUE <0.001 68.569 0.004 69.785 0.005  Total omega-6 0.000 0.996 TRUE <0.001 71.065 0.013 71.065 0.017  Ratio of omega-6 to omega-3 −0.021 0.127 TRUE <0.001 38.056 0.046 41.847 0.025  PUFA −0.004 0.806 TRUE <0.001 71.518 0.009 71.613 0.012  Ratio of PUFA to total fatty acids −0.013 0.379 TRUE <0.001 52.044 0.051 53.160 0.052 Premenstrual syndrome  DHA 0.020 0.405 TRUE <0.001 25.607 0.781 26.318 0.789  Linoleic acid 0.056 0.079 TRUE <0.001 39.048 0.468 42.311 0.372  Total omega-3 0.029 0.179 TRUE <0.001 39.065 0.557 40.931 0.518  Total omega-6 0.070 0.014 TRUE <0.001 38.704 0.800 45.249 0.586  Ratio of omega-6 to omega-3 0.004 0.878 TRUE <0.001 13.940 0.963 13.965 0.973  PUFA 0.048 0.094 TRUE <0.001 50.093 0.314 53.270 0.246  Ratio of PUFA to total fatty acids 0.012 0.655 TRUE <0.001 29.976 0.787 30.178 0.813 Premature ovarian insufficiency  DHA 0.001 0.971 TRUE <0.001 27.037 0.954 27.038 0.964  Linoleic acid 0.025 0.494 TRUE <0.001 45.547 0.408 46.039 0.429  Total omega-3 −0.003 0.912 TRUE <0.001 28.370 0.986 28.383 0.989  Total omega-6 −0.004 0.900 TRUE <0.001 45.908 0.804 45.924 0.829  Ratio of omega-6 to omega-3 0.013 0.655 TRUE <0.001 18.478 0.913 18.682 0.929  PUFA −0.023 0.440 TRUE <0.001 41.719 0.868 42.324 0.875  Ratio of PUFA to total fatty acids 0.000 0.990 TRUE <0.001 51.432 0.177 51.432 0.206 Note : Bold text indicates outcome disease categories. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; IVW, inverse-variance weighted. The Pleiotropic and Heterogeneity Test Note : Bold text indicates outcome disease categories. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; IVW, inverse-variance weighted. Subsequently, we focused on outlier SNP detection as detailed in Table 2 . In the MR-PRESSO and Radial MR tests, several outlier SNPs (eg., rs115478735) were identified. Where necessary, these outlier SNPs were removed to correct for potential horizontal pleiotropy. Importantly, the corrected analyses generally did not drastically alter the magnitude or direction of the causal estimates. For example, the causal relationships between Linoleic acid, Total omega-6, and PUFA with Endometriosis remained statistically insignificant even after the exclusion of the outlier SNPs (as shown in Table 1 and Table 2 ), thereby supporting the robustness of the main finding. Table 2 The MR-PRESSO and Radial Test in Forward MR Variables MR-PRESSO MR-PRESSO (Outlier Corrected) Radial MR Radial MR (Outlier Corrected) OR (95% CI) P OR (95% CI) P OR (95% CI) P OR (95% CI) P Endometriosis  DHA 0.857 (0.741–0.991) 0.043 0.997 (0.907–1.096) 0.951  Linoleic acid 1.127 (1.000–1.270) 0.057 1.127 (1.000–1.270) 0.051 1.097 (0.993–1.213) 0.069  Total omega-3 1.123 (1.001–1.259) 0.052 1.056 (0.975–1.143) 0.182  Total omega-6 0.997 (0.907–1.096) 0.951 1.123 (1.001–1.259) 0.047 1.096 (0.992–1.210) 0.071  Ratio of omega-6 to omega-3 1.056 (0.975–1.143) 0.188 1.089 (0.980–1.210) 0.111  PUFA 1.089 (0.980–1.210) 0.117 0.960 (0.879–1.048) 0.362 1.070 (0.977–1.171) 0.144  Ratio of PUFA to total fatty acids 0.960 (0.879–1.048) 0.370 0.857 (0.741–0.991) 0.037 Female infertility  DHA 0.972 (0.873–1.082) 0.606 0.972 (0.873–1.082) 0.603  Linoleic acid 1.059 (0.953–1.176) 0.293 1.155 (1.042–1.281) 0.006  Total omega-3 0.998 (0.909–1.096) 0.967 0.998 (0.909–1.096) 0.967  Total omega-6 1.109 (0.992–1.239) 0.073 1.109 (0.992–1.239) 0.068  Ratio of omega-6 to omega-3 0.870 (0.749–1.011) 0.077 1.059 (0.953–1.176) 0.288  PUFA 1.155 (1.042–1.281) 0.009 1.039 (0.970–1.113) 0.276  Ratio of PUFA to total fatty acids 1.039 (0.970–1.113) 0.285 0.870 (0.749–1.011) 0.070 Polycystic ovarian syndrome  DHA 0.765 (0.560–1.047) 0.102 0.957 (0.773–1.184) 0.684  Linoleic acid 1.214 (0.940–1.567) 0.146 1.214 (0.940–1.567) 0.138  Total omega-3 1.121 (0.867–1.450) 0.389 1.016 (0.841–1.226) 0.872  Total omega-6 1.071 (0.843–1.360) 0.578 1.121 (0.867–1.450) 0.384  Ratio of omega-6 to omega-3 0.995 (0.825–1.200) 0.958 1.071 (0.843–1.360) 0.575  PUFA 1.016 (0.841–1.226) 0.873 0.995 (0.825–1.200) 0.958  Ratio of PUFA to total fatty acids 0.957 (0.773–1.184) 0.687 0.765 (0.560–1.047) 0.094 Premenstrual syndrome  DHA 0.779 (0.460–1.322) 0.361 1.243 (0.876–1.763) 0.223  Linoleic acid 0.937 (0.564–1.556) 0.802 0.937 (0.564–1.556) 0.801  Total omega-3 1.376 (0.869–2.179) 0.18 1.294 (0.937–1.786) 0.118  Total omega-6 1.302 (0.820–2.068) 0.269 1.376 (0.869–2.179) 0.173  Ratio of omega-6 to omega-3 0.833 (0.654–1.061) 0.15 1.302 (0.820–2.067) 0.263  PUFA 1.294 (0.937–1.786) 0.125 0.833 (0.654–1.061) 0.138  Ratio of PUFA to total fatty acids 1.243 (0.876–1.763) 0.232 0.779 (0.459–1.322) 0.355 Premature ovarian insufficiency  DHA 0.728 (0.364–1.458) 0.375 0.765 (0.529–1.108) 0.156  Linoleic acid 0.765 (0.529–1.108) 0.163 1.458 (0.815–2.609) 0.204  Total omega-3 0.855 (0.637–1.148) 0.302 0.855 (0.637–1.148) 0.297  Total omega-6 1.458 (0.815–2.609) 0.211 1.534 (0.930–2.528) 0.094  Ratio of omega-6 to omega-3 1.533 (0.930–2.528) 0.099 1.040 (0.668–1.620) 0.862  PUFA 1.040 (0.668–1.620) 0.863 1.279 (0.934–1.752) 0.125  Ratio of PUFA to total fatty acids 1.279 (0.934–1.752) 0.136 0.728 (0.364–1.458) 0.370 Note : Bold text indicates outcome disease categories. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; MR-PRESSO, Mendelian randomization pleiotropy residual sum and outlier. The MR-PRESSO and Radial Test in Forward MR Note : Bold text indicates outcome disease categories. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; MR-PRESSO, Mendelian randomization pleiotropy residual sum and outlier. The primary MR analyses, based on IVW estimates (both fixed-effects and multiplicative random-effects), evaluated the relationships between PUFA-related exposures on various female endocrine disorders ( Table 3 ). The visualization of Mendelian randomization showed stable results, further confirming the reliability of our study findings. Graphical representations, including forest plots ( Figure 2 ) and scatter plots ( Figure 3 ), were used to illustrate these associations. Table 3 The MR Results of Exposures and Outcomes Variables nSNPs Fixed Effects Multiplicative Random Effects OR (95% CI) P OR (95% CI) P Endometriosis  DHA 43 0.997 (0.911–1.092) 0.949 0.997 (0.907–1.096) 0.951  Linoleic acid 46 1.127 (1.006–1.262) 0.039 1.127 (1.000–1.270) 0.051  Total omega-3 49 1.056 (0.979–1.138) 0.158 1.056 (0.975–1.143) 0.182  Total omega-6 57 1.123 (1.007–1.252) 0.037 1.123 (1.001–1.259) 0.047  Ratio of omega-6 to omega-3 30 0.960 (0.889–1.037) 0.296 0.960 (0.879–1.048) 0.362  PUFA 55 1.089 (0.987–1.202) 0.089 1.089 (0.980–1.210) 0.111  Ratio of PUFA to total fatty acids 45 0.857 (0.755–0.972) 0.017 0.857 (0.741–0.991) 0.037 Female infertility  DHA 43 0.972 (0.882–1.071) 0.564 0.972 (0.873–1.082) 0.603  Linoleic acid 46 1.155 (1.024–1.304) 0.019 1.155 (1.042–1.281) 0.006  Total omega-3 49 0.998 (0.921–1.081) 0.962 0.998 (0.909–1.096) 0.967  Total omega-6 57 1.109 (0.987–1.246) 0.082 1.109 (0.992–1.239) 0.068  Ratio of omega-6 to omega-3 30 1.039 (0.957–1.128) 0.362 1.039 (0.970–1.113) 0.276  PUFA 55 1.059 (0.953–1.176) 0.289 1.059 (0.953–1.176) 0.288  Ratio of PUFA to total fatty acids 45 0.870 (0.760–0.996) 0.044 0.870 (0.749–1.011) 0.070 Polycystic ovarian syndrome  DHA 34 0.957 (0.803–1.140) 0.620 0.957 (0.773–1.184) 0.684  Linoleic acid 41 1.214 (0.974–1.512) 0.085 1.214 (0.940–1.567) 0.138  Total omega-3 43 1.016 (0.878–1.175) 0.836 1.016 (0.841–1.226) 0.872  Total omega-6 49 1.121 (0.907–1.385) 0.290 1.121 (0.867–1.450) 0.384  Ratio of omega-6 to omega-3 27 0.995 (0.858–1.153) 0.947 0.995 (0.825–1.200) 0.958  PUFA 48 1.071 (0.882–1.300) 0.489 1.071 (0.843–1.360) 0.575  Ratio of PUFA to total fatty acids 45 0.765 (0.587–0.997) 0.048 0.765 (0.560–1.047) 0.094 Premenstrual syndrome  DHA 34 1.243 (0.840–1.838) 0.277 1.243 (0.876–1.763) 0.223  Linoleic acid 41 0.937 (0.572–1.534) 0.795 0.937 (0.564–1.556) 0.801  Total omega-3 43 1.294 (0.933–1.794) 0.123 1.294 (0.937–1.786) 0.118  Total omega-6 49 1.376 (0.857–2.209) 0.186 1.376 (0.869–2.179) 0.173  Ratio of omega-6 to omega-3 27 0.833 (0.598–1.158) 0.277 0.833 (0.654–1.061) 0.138  PUFA 48 1.302 (0.843–2.010) 0.234 1.302 (0.820–2.068) 0.263  Ratio of PUFA to total fatty acids 39 0.779 (0.431–1.410) 0.410 0.779 (0.460–1.322) 0.355 Premature ovarian insufficiency  DHA 43 0.765 (0.482–1.213) 0.255 0.765 (0.529–1.108) 0.156  Linoleic acid 46 1.458 (0.820–2.592) 0.199 1.458 (0.815–2.609) 0.204  Total omega-3 49 0.855 (0.583–1.254) 0.422 0.855 (0.637–1.148) 0.297  Total omega-6 57 1.533 (0.883–2.663) 0.129 1.533 (0.930–2.528) 0.094  Ratio of omega-6 to omega-3 30 1.279 (0.864–1.893) 0.219 1.279 (0.934–1.752) 0.125  PUFA 55 1.040 (0.631–1.716) 0.878 1.040 (0.668–1.620) 0.862  Ratio of PUFA to total fatty acids 45 0.728 (0.383–1.384) 0.333 0.728 (0.364–1.458) 0.370 Notes : Bold text in the Variables column indicates outcome disease categories, and bold OR and P values indicate statistically significant associations with P < 0.05. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; SNP, single nucleotide polymorphism; OR, odds ratio; CI, confidence interval. Figure 2 Forest plot of the forward Mendelian randomization analysis using fixed-effect IVW and multiplicative random-effect IVW models. The left and right panels show estimates from the fixed-effect IVW and multiplicative random-effect IVW models, respectively. Dots represent OR estimates, horizontal bars represent 95% CIs, and the vertical dashed line indicates OR = 1. Orange and blue text indicate statistically significant risk and protective effects, respectively, while black text indicates non-significant associations. Bold text indicates grouped PUFA-related exposure categories, and indented non-bold text indicates the corresponding outcome traits. Two forest plots of odds ratios for polyunsaturated fatty acid exposures across female endocrine outcomes. The text analyzes the link between fatty acids and conditions like endometriosis, infertility, PCOS, PMS and premature ovarian insufficiency using forest plots showing odds ratios (OR) and 95% confidence intervals (CI). Key findings: 1. **Endometriosis**: Linoleic acid slightly raises risk (OR ~1.127), while PUFA to total fatty acids ratio is protective (OR ~0.857). 2. **Female Infertility**: Linoleic acid increases risk (OR ~1.155), with PUFA ratio showing protection (OR ~0.870). 3. **PCOS**: Linoleic acid raises risk (OR ~1.214), PUFA ratio is protective (OR ~0.765). 4. **PMS**: Total omega-6 increases risk (OR ~1.376), omega-6 to omega-3 ratio is protective (OR ~0.833). 5. **Premature Ovarian Insufficiency**: Total omega-6 raises risk (OR ~1.533), PUFA ratio is protective (OR ~0.728). Each condition′s data is shown with OR dots and horizontal CI bars, indicating estimate precision. Abbreviations : IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; CI, confidence interval; SNP, single nucleotide polymorphism; DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid. Figure 3 The scatter plots for different Mendelian randomization (MR) methods show consistent β values. ( A ) Total omega-6 to Endometriosis. ( B ) Linoleic acid to Endometriosis. ( C ) Linoleic acid to Female infertility. ( D ) Ratio of PUFA to total fatty acids to Polycystic ovarian syndrome. ( E ) Ratio of PUFA to total fatty acids to Endometriosis. ( F ) Ratio of PUFA to total fatty acids to Female infertility. Six scatter plots of Mendelian randomization slopes, showing mostly weak positive or negative trends. The image A showing a scatter plot with error bars and four fitted lines under the heading “MR Test”. The x-axis label is “SNP effect on Total omega-6” with ticks at 0.1, 0.2, 0.3. The y-axis label is “SNP effect on Endometriosis” with ticks at minus 0.3, minus 0.2, minus 0.1, 0.0, 0.1. Points cluster near x about 0.0 to 0.1 and y about 0.0. All four lines slope slightly upward. The image B showing a scatter plot with error bars and four fitted lines under “MR Test”. The x-axis label is “SNP effect on Linoleic acid” with ticks at 0.1 and 0.2. The y-axis label is “SNP effect on Endometriosis” with ticks at minus 0.2, 0.0, 0.2. Points cluster near x about 0.0 to 0.1 and y about 0.0. All four lines slope slightly upward. The image C showing a scatter plot with error bars and four fitted lines under “MR Test”. The x-axis label is “SNP effect on Linoleic acid” with ticks at 0.1 and 0.2. The y-axis label is “SNP effect on Female infertility” with ticks at 0.0, 0.2, 0.4. Points cluster near x about 0.0 to 0.1 and y about 0.0, with one point near y about 0.3. All four lines slope slightly upward. The image D showing a scatter plot with error bars and four fitted lines under “MR Test”. The x-axis label is “SNP effect on Ratio of PUFA to total fatty acids” with ticks at 0.05, 0.10, 0.15, 0.20. The y-axis label is “SNP effect on Polycystic ovarian syndrome” with ticks at minus 0.25, 0.00, 0.25. Points cluster near x about 0.02 to 0.08 and y near 0.0. All four lines slope slightly downward. The image E showing a scatter plot with error bars and four fitted lines under “MR Test”. The x-axis label is “SNP effect on Ratio of PUFA to total fatty acids” with ticks at 0.05, 0.10, 0.15, 0.20. The y-axis label is “SNP effect on Endometriosis” with ticks at minus 0.2, 0.0, 0.2. Points cluster near x about 0.02 to 0.08 and y near 0.0. All four lines slope slightly downward. The image F showing a scatter plot with error bars and four fitted lines under “MR Test”. The x-axis label is “SNP effect on Ratio of PUFA to total fatty acids” with ticks at 0.05, 0.10, 0.15, 0.20. The y-axis label is “SNP effect on Female infertility” with ticks at minus 0.4, minus 0.2, 0.0, 0.2. Points cluster near x about 0.02 to 0.08 and y near 0.0, with one point near y about minus 0.3. All four lines slope slightly downward. The MR Results of Exposures and Outcomes Notes : Bold text in the Variables column indicates outcome disease categories, and bold OR and P values indicate statistically significant associations with P < 0.05. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; SNP, single nucleotide polymorphism; OR, odds ratio; CI, confidence interval. Forest plot of the forward Mendelian randomization analysis using fixed-effect IVW and multiplicative random-effect IVW models. The left and right panels show estimates from the fixed-effect IVW and multiplicative random-effect IVW models, respectively. Dots represent OR estimates, horizontal bars represent 95% CIs, and the vertical dashed line indicates OR = 1. Orange and blue text indicate statistically significant risk and protective effects, respectively, while black text indicates non-significant associations. Bold text indicates grouped PUFA-related exposure categories, and indented non-bold text indicates the corresponding outcome traits. The scatter plots for different Mendelian randomization (MR) methods show consistent β values. ( A ) Total omega-6 to Endometriosis. ( B ) Linoleic acid to Endometriosis. ( C ) Linoleic acid to Female infertility. ( D ) Ratio of PUFA to total fatty acids to Polycystic ovarian syndrome. ( E ) Ratio of PUFA to total fatty acids to Endometriosis. ( F ) Ratio of PUFA to total fatty acids to Female infertility. A notably consistent finding was the positive association of genetically predicted linoleic acid levels with both endometriosis and female infertility. In the fixed-effects IVW analysis, linoleic acid exhibited an odds ratio (OR) of approximately 1.127 for endometriosis and 1.155 for female infertility ( Figure 2 ); the random-effects models yielded similar or slightly stronger estimates. These results suggest that elevated circulating linoleic acid may increase susceptibility to these gynecological conditions ( Figures 3B and C ). Total omega-6 also demonstrated a significant positive causal association with endometriosis, with an OR near 1.123 in the fixed-effects IVW analysis ( Figure 2 ). The random-effects analysis was marginally significant as well, indicating that higher total omega-6 levels might contribute to an increased risk of endometriosis ( Figure 3A ). In contrast, an intriguing inverse association was observed for the ratio of PUFA to total fatty acids (TFAs). This ratio appeared protective against endometriosis, female infertility, and polycystic ovarian syndrome. Specifically, for endometriosis, the fixed-effects IVW analysis showed an OR of roughly 0.857 ( Figure 2 ), with similar trends observed under random-effects modeling. For female infertility and PCOS, protective effects were apparent under fixed-effects, although these associations were less robust in the random-effects models ( Figure 3D–F ) visually depict how higher values of this ratio align with lower disease risk). Additionally, leave-one-out analysis validated the stability of the results, with no single SNP significantly altering the causal estimates (see Figure S1 – S6 ). These sensitivity analyses collectively affirm the robustness of our MR results. Furthermore, no definitive causal relationships were consistently identified for DHA, total omega-3, the ratio of omega-6 to omega-3, or total PUFA with respect to most outcomes (including premenstrual syndrome and premature ovarian insufficiency). Although some secondary or sensitivity analyses (eg., MR-PRESSO results for DHA and endometriosis) suggested minor trends, these findings were not supported by the primary IVW analyses. Overall, the IVW results indicate that increased circulating levels of linoleic acid and total omega-6 may elevate the risk of endometriosis and, in the case of linoleic acid, female infertility, whereas a higher ratio of PUFA to total fatty acids may confer a protective effect against endometriosis, female infertility, and polycystic ovarian syndrome. Additional analyses using weighted median and weighted mode methods are provided in supplementary materials ( Table S3 ). We additionally tested whether reproductive endocrine disorders—endometriosis, female infertility, PCOS, premenstrual syndrome, and POI—causally affect circulating PUFA traits. In these reverse models, the MR Egger intercept tests generally indicated minimal horizontal pleiotropy and the Cochran’s Q tests suggested little heterogeneity. However, for PCOS, evidence of horizontal pleiotropy was detected by the MR-Egger intercept test for SNPs associated with circulating PUFA, the ratio of PUFA to total fatty acids, total omega-6, and linoleic acid, whereas Cochran’s Q test continued to show no significant heterogeneity ( Table S4 ). Among all reverse analyses, polycystic ovarian syndrome showed the strongest potential to influence certain PUFA measures. Specifically, genetic liability to PCOS was associated with decreased levels of the ratio of PUFA to total fatty acids, total omega-3, and DHA, and a potential increase in the ratio of omega-6 to omega-3. Although the effect sizes were modest, the directions of these associations were consistent across both fixed-effects and multiplicative random-effects IVW models ( Table 4 ) and Figure 4 visually depicts an inverse association for PCOS with the ratio of PUFA to total fatty acids, total omega-3, and DHA (all below unity), as well as a positive shift in the ratio of omega-6 to omega-3 in certain analyses. Table 4 The Reverse MR Results of Exposures and Outcomes Variables nSNPs Fixed Effects Multiplicative Random Effects OR (95% CI) P OR (95% CI) P PUFA Endometriosis 11 1.011 (0.993–1.031) 0.233 1.011 (0.993–1.030) 0.213 Polycystic ovarian syndrome 196 0.998 (0.995–1.001) 0.121 0.998 (0.996–1.000) 0.051 Ratio of PUFA to total fatty acids Endometriosis 11 0.987 (0.969–1.005) 0.166 0.987 (0.963–1.012) 0.303 Polycystic ovarian syndrome 196 0.990 (0.988–0.993) 0.000 0.990 (0.988–0.993) 0.000 Total omega-3 Endometriosis 11 1.009 (0.991–1.028) 0.318 1.009 (0.988–1.032) 0.398 Polycystic ovarian syndrome 196 0.997 (0.994–1.000) 0.032 0.997 (0.995–0.999) 0.006 DHA Endometriosis 11 1.004 (0.985–1.022) 0.692 1.004 (0.980–1.028) 0.765 Polycystic ovarian syndrome 196 0.989 (0.987–0.992) 0.000 0.989 (0.987–0.991) 0.000 Total omega-6 Endometriosis 11 1.010 (0.991–1.029) 0.290 1.010 (0.991–1.030) 0.300 Polycystic ovarian syndrome 196 0.999 (0.996–1.001) 0.319 0.999 (0.996–1.001) 0.225 Linoleic acid Endometriosis 11 1.010 (0.991–1.029) 0.307 1.010 (0.988–1.033) 0.387 Polycystic ovarian syndrome 196 1.000 (0.997–1.002) 0.788 1.000 (0.998–1.002) 0.707 Ratio of omega-6 to omega-3 Endometriosis 11 0.995 (0.976–1.013) 0.567 0.995 (0.971–1.019) 0.664 Polycystic ovarian syndrome 196 1.003 (1.001–1.006) 0.015 1.003 (1.001–1.006) 0.001 Notes : Bold text in the Variables column indicates grouped PUFA-related traits, and bold OR and P values indicate statistically significant associations with P < 0.05. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; SNP, single nucleotide polymorphism; OR, odds ratio; CI, confidence interval. Figure 4 Forest plot of the reverse Mendelian randomization analysis using fixed-effect IVW and multiplicative random-effect IVW models. The left and right panels show estimates from the fixed-effect IVW and multiplicative random-effect IVW models, respectively. Dots represent OR estimates, horizontal bars represent 95% CIs, and the vertical dashed line indicates OR = 1. Orange and blue text indicate statistically significant positive and negative associations, respectively, while black text indicates non-significant associations. Bold text indicates grouped reproductive disorder categories, and indented non-bold text indicates the corresponding PUFA-related traits. Two forest plots comparing odds ratio estimates for reproductive disorders across PUFA-related traits. Two side-by-side forest plots with a shared table of results. The image A showing fixed-effect IVW results. Column headers read: Variables, fix nSNPs, fix OR(95 percentCI), fix outcome. The x-axis label is OR (unitless), with tick labels 0.97, 1 and 1.03 and a vertical dashed reference line at OR equals 1. Rows list: PUFA: Endometriosis, fix nSNPs 11, OR 1.011 (0.993 to 1.031). Polycystic ovarian syndrome, fix nSNPs 196, OR 0.998 (0.995 to 1.001). Ratio of PUFA to total fatty acids: Endometriosis, 11, OR 0.987 (0.969 to 1.005). Polycystic ovarian syndrome, 196, OR 0.990 (0.988 to 0.993). Total omega-3: Endometriosis, 11, OR 1.009 (0.991 to 1.028). Polycystic ovarian syndrome, 196, OR 0.997 (0.994 to 1.000). DHA: Endometriosis, 11, OR 1.004 (0.985 to 1.022). Polycystic ovarian syndrome, 196, OR 0.989 (0.987 to 0.992). Total omega-6: Endometriosis, 11, OR 1.010 (0.991 to 1.029). Polycystic ovarian syndrome, 196, OR 0.999 (0.996 to 1.001). Linoleic acid: Endometriosis, 11, OR 1.010 (0.991 to 1.029). Polycystic ovarian syndrome, 196, OR 1.000 (0.997 to 1.002). Ratio of omega-6 to omega-3: Endometriosis, 11, OR 0.995 (0.976 to 1.013). Polycystic ovarian syndrome, 196, OR 1.003 (1.001 to 1.006). The image B showing multiplicative random-effect IVW results. Column headers read: mul nSNPs, mul OR(95 percentCI), mul outcome. The x-axis label is OR (unitless), with tick labels 0.97, 1 and 1.03 and a vertical dashed reference line at OR equals 1. Rows list: PUFA: Endometriosis, mul nSNPs 11, OR 1.011 (0.993 to 1.030). Polycystic ovarian syndrome, 196, OR 0.998 (0.996 to 1.000). Ratio of PUFA to total fatty acids: Endometriosis, 11, OR 0.987 (0.963 to 1.012). Polycystic ovarian syndrome, 196, OR 0.990 (0.988 to 0.993). Total omega-3: Endometriosis, 11, OR 1.009 (0.988 to 1.032). Polycystic ovarian syndrome, 196, OR 0.997 (0.995 to 0.999). DHA: Endometriosis, 11, OR 1.004 (0.980 to 1.028). Polycystic ovarian syndrome, 196, OR 0.989 (0.987 to 0.991). Total omega-6: Endometriosis, 11, OR 1.010 (0.991 to 1.030). Polycystic ovarian syndrome, 196, OR 0.999 (0.996 to 1.001). Linoleic acid: Endometriosis, 11, OR 1.010 (0.988 to 1.033). Polycystic ovarian syndrome, 196, OR 1.000 (0.998 to 1.002). Ratio of omega-6 to omega-3: Endometriosis, 11, OR 0.995 (0.971 to 1.019). Polycystic ovarian syndrome, 196, OR 1.003 (1.001 to 1.006). Abbreviations : IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; CI, confidence interval; SNP, single nucleotide polymorphism; DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid. The Reverse MR Results of Exposures and Outcomes Notes : Bold text in the Variables column indicates grouped PUFA-related traits, and bold OR and P values indicate statistically significant associations with P < 0.05. Abbreviations : DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid; MR, Mendelian randomization; SNP, single nucleotide polymorphism; OR, odds ratio; CI, confidence interval. Forest plot of the reverse Mendelian randomization analysis using fixed-effect IVW and multiplicative random-effect IVW models. The left and right panels show estimates from the fixed-effect IVW and multiplicative random-effect IVW models, respectively. Dots represent OR estimates, horizontal bars represent 95% CIs, and the vertical dashed line indicates OR = 1. Orange and blue text indicate statistically significant positive and negative associations, respectively, while black text indicates non-significant associations. Bold text indicates grouped reproductive disorder categories, and indented non-bold text indicates the corresponding PUFA-related traits. In contrast, endometriosis, female infertility, premenstrual syndrome, and POI did not exhibit robust causal effects on the measured PUFA phenotypes. Across multiple tests, most estimates were centered around null values, providing no strong evidence that these conditions directly alter fatty acid profiles in circulation. Graphical representations of the forward MR results ( Figure 2 ) illustrate that the slope of the fitted lines—whether estimated via IVW or MR-Egger—remains consistent with the statistical findings. Our analysis indicates that elevated levels of linoleic acid or total omega-6 are implicated in an increased risk of several reproductive disorders. Conversely, a higher PUFA-to-TFA ratio is linked to a protective effect against these conditions.In contrast, Figure 4 , which presents the reverse MR scenario through forest plots, confirms that PCOS is consistently linked with lower levels of beneficial PUFA fractions, including DHA, total omega-3, and the ratio of PUFA to total fatty acids, while also showing an increased ratio of omega-6 to omega-3. Collectively, these analyses indicate that elevated circulating linoleic acid and total omega-6 may raise the risk of endometriosis—and, in the case of linoleic acid, female infertility—while a higher ratio of PUFAs to TFAs may protect against endometriosis, female infertility, and PCOS. Conversely, reverse MR analyses suggest that genetic liability to PCOS may reduce beneficial PUFA-related measures and shift the balance towards a more pro-inflammatory state by increasing the ratio of omega-6 to omega-3. These findings underscore the importance of monitoring specific fatty acid subtypes in women’s health and suggest that interventions targeting the balance between omega-6 and omega-3 could potentially mitigate the burden of endometriosis, infertility, and other reproductive conditions in susceptible populations.

This study employed a two-sample Mendelian randomization (MR) design to examine the causal effects of circulating polyunsaturated fatty acids (PUFAs) on female reproductive endocrine disorders ( Figure 1 ). Figure 1 Directed acyclic graph illustrating the bidirectional Mendelian randomization framework between circulating polyunsaturated fatty acid-related traits and female reproductive endocrine disorders. Solid arrows indicate assumed causal pathways, and the double-headed arrow indicates the bidirectional MR design. Red cross symbols indicate that genetic instruments should not be associated with confounders or directly affect outcomes except through exposures. Bold text denotes the main analytical components. Diagram of genetic instruments, exposures, confounders and outcomes in MR analysis. The diagram shows Mendelian randomization (MR) analysis, highlighting genetic instruments, exposures, confounders and outcomes. Genetic instruments influence exposures and outcomes, while confounders affect both. Exposures include DHA, linoleic acid, omega-3, omega-6, PUFA and their ratios, each valued at 115,006. Outcomes are endometriosis (77,257), female infertility (75,450), ovarian syndrome (243,907), premenstrual syndrome (119,843) and premature ovarian insufficiency (118,228). MR methods for causal inference include IVW, MR-Egger, weighted mode, weighted median, radial MR, MR-PRESSO and leave-one-out analysis. Red crosses indicate genetic instruments should not be linked to confounders or directly impact outcomes except via exposures. Abbreviations : IVW, inverse-variance weighted; MR, Mendelian randomization; DHA, docosahexaenoic acid; PUFA, polyunsaturated fatty acid. Directed acyclic graph illustrating the bidirectional Mendelian randomization framework between circulating polyunsaturated fatty acid-related traits and female reproductive endocrine disorders. Solid arrows indicate assumed causal pathways, and the double-headed arrow indicates the bidirectional MR design. Red cross symbols indicate that genetic instruments should not be associated with confounders or directly affect outcomes except through exposures. Bold text denotes the main analytical components. By employing genetic variants as instrumental variables (IVs), the MR approach provides a robust method to infer causality, addressing limitations inherent in observational studies, such as residual confounding and reverse causation. The validity of MR relies on three key assumptions. First, the relevance assumption requires that the selected IVs are strongly associated with the exposure of interest, ensuring that genetic variants significantly predict circulating PUFA levels. Second, the independence assumption mandates that IVs are not associated with any confounders of the exposure-outcome relationship, which is generally supported by the random allocation of alleles during meiosis. Third, the exclusion restriction assumption specifies that IVs influence the outcome only through their effects on the exposure, with no alternative pathways such as horizontal pleiotropy. These assumptions were rigorously evaluated and validated to ensure reliable causal inference. Genetic associations for circulating PUFA levels were sourced from a large-scale genome-wide association study (GWAS) of 115,006 individuals of European ancestry. This GWAS dataset provided comprehensive coverage of PUFA-related traits, including docosahexaenoic acid (DHA), linoleic acid, total omega-3 fatty acids, total omega-6 fatty acids, the ratio of omega-6 to omega-3 fatty acids, total PUFAs, and the ratio of PUFAs to total fatty acids. These data were sourced from the European Bioinformatics Institute (EBI) database, with relevant summary statistics published under GWAS Catalog IDs ranging from ebi-a-GCST90092816 to ebi-a-GCST90092941. The high-quality data from this large cohort provided a robust foundation for exploring the genetic determinants of PUFA levels and their subsequent effects on health outcomes. The outcome data were derived from GWAS datasets representing European ancestry populations from the FinnGen biobank analysis round 5. Specifically, endometriosis was examined in 77,257 participants (of whom 8,288 were cases), 32 female infertility in 75,450 participants, polycystic ovarian syndrome in 220,609 individuals, premenstrual syndrome in 111,878 participants, and premature ovarian insufficiency in 118,228 participants. These datasets were selected for their large sample sizes, comprehensive genomic coverage, and rigorous phenotypic characterization, which together maximize statistical power and enhance the reliability of causal estimates. All data are publicly available at https://gwas.mrcieu.ac.uk/ . Instrumental variables were selected from the PUFA GWAS based on genome-wide significance thresholds ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$P < 5 \times {10^{ - 8}}$\end{document} ) to ensure strong associations with PUFA levels, satisfying the relevance assumption. To maintain independence between selected variants, clumping procedures were applied to remove linkage disequilibrium (LD) using an \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${r^2}$\end{document} threshold of 0.001 within a 10,000 kilobase window. Palindromic single nucleotide polymorphisms (SNPs) with intermediate allele frequencies were excluded to minimize strand ambiguity and ensure consistency in allele alignment across datasets. These rigorous criteria ensured that the selected SNPs were robustly associated with circulating PUFA levels while minimizing potential bias introduced by LD or technical errors. The potential influence of horizontal pleiotropy, where IVs affect outcomes through pathways independent of the exposure, was assessed using multiple approaches. MR-Egger regression was employed to detect directional pleiotropy, with the intercept term providing a quantitative measure of deviation from the exclusion restriction assumption. A significant intercept term ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$P < 0.05$\end{document} ) indicated the presence of pleiotropy. To address this, sensitivity analyses were conducted using methods robust to pleiotropy, including the weighted mode and weighted median approaches, which provide valid estimates even when a portion of the IVs are invalid. Additionally, MR-PRESSO (Pleiotropy Residual Sum and Outlier) was used to detect and correct for outliers that might bias causal estimates. Harmonization of SNPs across exposure and outcome datasets ensured consistent alignment of alleles, further reducing the risk of pleiotropic bias. The strength of the selected IVs was evaluated using the \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$F$\end{document} -statistic and the proportion of variance in the exposure explained by the IVs ( \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${R^2}$\end{document} ). 33 The \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$F$\end{document} -statistic was calculated using the formula: (1) \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$$F = {{{R^2}\left({n - 1 - k} \right)} \over {\left({1 - {R^2}} \right)k}}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$n$\end{document} is the sample size, \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$k$\end{document} is the number of IVs, and \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}${R^2}$\end{document} represents the proportion of exposure variance explained by the IVs. 34 An \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$F$\end{document} -statistic exceeding 10 was used to define strong instruments, minimizing the risk of weak instrument bias and ensuring robust causal inference. Causal effects were estimated using the inverse variance-weighted (IVW) method, the primary approach for two-sample MR analyses. Both multiplicative random-effects and fixed-effects models were applied, with random-effects results prioritized in the presence of heterogeneity. Sensitivity analyses complemented the primary IVW approach, including the weighted median method, which provides robust estimates when up to 50% of IVs are invalid, and the MR-Egger regression, which accounts for pleiotropy under the Instrument Strength Independent of Direct Effect (InSIDE) assumption. MR-PRESSO was used to identify and correct for outlier SNPs, while Radial MR was implemented to detect influential SNPs and assess small-sample bias. To further evaluate the robustness of causal estimates, a leave-one-out analysis was conducted by systematically excluding individual SNPs from the analysis. Heterogeneity among SNP-specific causal estimates was assessed using Cochran’s \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$Q$\end{document} statistic, which quantifies the deviation of individual SNP effects from the overall IVW estimate. Significant heterogeneity, indicated by \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$P < 0.05$\end{document} , prompted the use of random-effects models, which account for between-SNP variability. The presence of heterogeneity was considered when interpreting results, as it may reflect biological differences in SNP effects or potential violations of MR assumptions. To examine the potential for reverse causation, the MR-Steiger test was conducted, evaluating whether the genetic variants explained more variance in the exposure than in the outcome. This approach ensured the validity of causal directionality, distinguishing exposure-driven effects from outcome-driven genetic associations. Sensitivity analyses included the consistency of results across multiple MR methods, such as IVW, MR-Egger, weighted median, and weighted mode. Results were considered robust if causal estimates were consistent in direction and magnitude across methods. Statistical significance was defined at a two-sided \documentclass[12pt]{minimal} \usepackage{wasysym} \usepackage[substack]{amsmath} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage[mathscr]{eucal} \usepackage{mathrsfs} \DeclareFontFamily{T1}{linotext}{} \DeclareFontShape{T1}{linotext}{m}{n} {linotext }{} \DeclareSymbolFont{linotext}{T1}{linotext}{m}{n} \DeclareSymbolFontAlphabet{\mathLINOTEXT}{linotext} \begin{document}$\alpha = 0.05$\end{document} , with findings contextualized within biological plausibility and prior literature. All analyses were conducted using the TwoSampleMR R package and related statistical tools, ensuring transparency and reproducibility in the analytical process.

In summary, our two-sample Mendelian randomization study provides genetic evidence supporting a potential causal association between circulating polyunsaturated fatty acid profiles and female reproductive endocrine disorders. Genetically predicted higher circulating linoleic acid and total omega-6 levels were associated with increased risks of endometriosis and female infertility, whereas a higher genetically predicted ratio of PUFAs to total fatty acids was associated with lower risks of these conditions and PCOS. Reverse MR analyses further suggested that PCOS may influence circulating lipid profiles, which may partly reflect altered endogenous lipid metabolism and inflammatory regulation. Therefore, our findings suggest that PUFA-related metabolic pathways may contribute to the pathophysiology of female reproductive endocrine disorders, while the potential benefits of dietary or therapeutic modulation of PUFA profiles require further validation in mechanistic studies, randomized trials, and diverse populations.

This study applied a two-sample Mendelian randomization (MR) framework to examine the causal relationships between circulating PUFA traits and female reproductive endocrine disorders—including endometriosis, female infertility, PCOS, PMS and POI. By leveraging both forward and reverse MR analyses, our study addresses the inherent limitations of traditional observational studies, such as unmeasured confounding and reverse causation, thereby strengthening the evidence for causal conclusions. In the forward MR analyses, elevated levels of linoleic acid and total omega-6 were positively associated with endometriosis and female infertility, with linoleic acid showing the strongest and most consistent effects. It could be explained by the role of omega-6-related lipid metabolism in reproductive tract inflammation. 35 Linoleic acid is the parent omega-6 PUFA and can be metabolically connected to downstream omega-6 pathways, including arachidonic acid-derived eicosanoid production. 36 Arachidonic acid serves as a precursor for prostaglandins and leukotrienes, which are central mediators of inflammatory signaling. 36 On the other hand, prostaglandins and leukotrienes could be directly involved in reproductive physiology and pathology. In endometriosis, endometriotic lesions show increased COX-2 expression and COX-2-derived PGE2 biosynthesis compared with normal endometrium. 37 COX-2/PGE2 axis promotes inflammatory activation, lesion survival, angiogenesis, pain sensitization, and local estrogen production, thereby supporting the inflammatory microenvironment required for lesion persistence. 37 Leukotrienes, generated through the 5-lipoxygenase pathway, may further amplify inflammatory cell recruitment, vascular permeability, and peritoneal inflammatory responses. 36 For female infertility, PUFA-derived prostaglandin signaling may also influence reproductive physiology by regulating prostaglandin synthesis, steroid metabolism, ovulation, endometrial receptivity, and implantation. 35 Conversely, a higher ratio of PUFAs to TFAs appeared to be protective against endometriosis, female infertility, and PCOS. This protective effect likely reflects the anti-inflammatory role of omega-3 PUFAs, which counteract omega-6-mediated inflammation by producing mediators such as resolvins and protectins. 38 Notably, the ratio of omega-6 to omega-3, an established marker of inflammatory balance 39 did not show significant causal associations, suggesting that overall PUFA balance may be more critical than their relative proportions. Reverse MR analyses further revealed that genetic liability to PCOS might lead to systemic changes in lipid profiles, specifically by reducing total omega-3 and DHA levels and increasing the ratio of omega-6 to omega-3. Although the effect sizes were modest, these findings suggest that PCOS may shift the PUFA profile toward a more pro-inflammatory state, potentially exacerbating the condition. One possible explanation involves the FADS1–FADS2 desaturase pathway. The FADS gene cluster has been associated with PCOS susceptibility, and FADS2 encodes Δ6 desaturase, a key enzyme involved in the endogenous conversion of essential fatty acid precursors, including α-linolenic acid and linoleic acid, into their long-chain PUFA metabolites. 40 , 41 Because omega-3 and omega-6 precursors share desaturase and elongase enzymes, reduced FADS2 expression or activity could create a metabolic bottleneck that limits the conversion of α-linolenic acid toward long-chain omega-3 species, including DHA, while favoring a higher omega-6 to omega-3 balance under conditions of abundant omega-6 substrate availability. 42 , 43 In addition to the FADS1–FADS2 pathway, hormonal imbalance may also partly explain the reverse associations between PCOS and altered PUFA profiles. Female reproductive endocrine disorders are characterized by distinct endocrine disturbances, including hyperandrogenism and ovulatory dysfunction in PCOS, estrogen-dependent inflammatory activity in endometriosis, impaired follicular development in female infertility, cyclic ovarian steroid fluctuations in PMS, and estrogen–gonadotropin dysregulation in POI. 44–46 These hormonal abnormalities may influence lipid metabolism, inflammatory mediator production, and circulating PUFA levels. Moreover, PCOS and other inflammatory reproductive disorders are closely linked to immune-inflammatory dysregulation, including altered interleukin and cytokine signaling, which may further connect endocrine disturbance with PUFA-related inflammatory metabolism. 47–49 This study has several notable strengths. First, the use of both forward and reverse MR analyses minimizes biases from confounding and reverse causation, offering robust causal inferences. Second, the utilization of large-scale GWAS datasets provided comprehensive genetic coverage and statistical power, particularly for European populations. Third, extensive sensitivity analyses—including MR-Egger, weighted median, and MR-PRESSO—bolstered the credibility of our findings by addressing potential horizontal pleiotropy and weak instrument bias. However, certain limitations should be acknowledged. The absence of individual-level data restricted our ability to perform more detailed analyses on the causal relationships between specific circulating PUFA traits and reproductive disorders. Because this study was based on publicly available GWAS summary-level data, we could not directly control the original inclusion and exclusion criteria of the recruited participants, such as disease severity, age distribution, or other cohort-specific clinical characteristics. Also, similar studies in non-European and multi-ethnic populations are warranted for the validation. Meanwhile, as this study was based on summary-level genetic data, the exploration of specific biological mechanisms underlying the observed associations could be interesting in the future. Finally, because our study was based on genetic data from individuals of European descent, the findings may not be generalizable to other ethnic groups, which could have differing genetic architectures and dietary patterns. Overall, our findings have important implications for public health and clinical practice. The identification of elevated linoleic acid and total omega-6 levels as potential risk factors for endometriosis and infertility suggests that dietary interventions aimed at reducing omega-6 intake and promoting omega-3 consumption could be beneficial. Additionally, the protective association observed with a higher ratio of PUFAs to TFAs highlights the potential of maintaining a balanced PUFA profile to improve reproductive health. Future research should aim to validate these findings in more diverse populations and explore gene-diet interactions that may further influence these associations, as well as elucidate the underlying biological mechanisms through experimental studies.